Crop recommendation and yield prediction using machine learning algorithms

Sundari V, Anusree M, Swetha U and Divya Lakshmi R *

Department of Computer Science and Engineering, Meenakshi Sundararajan Engineering College, Anna University,
Chennai, India.

World Journal of Advanced Research and Reviews, 2022, 14(03), 452–459

Publication history: Received on 13 May 2022; revised on 16 June 2022; accepted on 18 June 2022

Article DOI: https://doi.org/10.30574/wjarr.2022.14.3.0581

Abstract
Agriculture is the foundation of many countries' economies, particularly in India and Tamil Nadu. The young generation
who are new to farming may confront the challenge of not understanding what to sow and what to reap benefit from.
This is a problem that has to be addressed, and it is one that we are addressing. Predicting the proper crop and
production will aid in making better decisions, reducing losses and managing the risk of price fluctuations. The existing
system is not deployed, unlike ours, which is done by applying classification and regression algorithms to calculate crop
type recommendations and yield predictions. Agricultural industries must use machine learning algorithms to
anticipate the crop from a given dataset. The supervised machine learning technique is used to analyse a dataset in
order to capture information from multiple sources, such as variable identification, uni-variate analysis, bi-variate and
multi-variate analysis, missing value treatments, and so on. A comparison of machine learning algorithms was
conducted in order to identify which algorithm was more accurate in predicting the best harvest. The results show that
the proposed machine learning algorithm technique has the best accuracy when comparing entropy calculation,
precision, Recall, F1 Score, Sensitivity, Specificity, and Entropy.

We have ensured that our proposed system accomplishes its job effectively by projecting the yield of practically all types
of crops grown in Tamil Nadu, relieving some of the burden from their shoulders as they enter a new business.

Keywords: Supervised Machine Learning Approach; Classification and Regression Models; Precision; Linear
Regression; Logistic Regression; Decision Tree and Random Forest

1. Introduction
Agriculture is an important source of income for many people in underdeveloped countries. Several technologies,
conditions, practices, and civilizations have all influenced agricultural expansion in recent years. Furthermore, the use
of information technology may alter the state of decision-making, allowing farmers to produce the best results. Data
mining techniques connected to agriculture are employed in the decision-making process. The process of extracting the
most important and relevant information from a large number of datasets is known as data mining. As agriculture
involves a variety of data, such as soil data, crop data, and weather data, we now employ a machine learning approach
designed for crop or plant yield prediction. Machine learning techniques are efficiently used to propose a crop
recommendation and yield prediction system.


Corresponding author: Divya Lakshmi R
Department of Computer Science and Engineering, Meenakshi Sundararajan Engineering College (Affiliated to Anna University)
Chennai, India.
Copyright © 2022 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution Liscense 4.0.
 World Journal of Advanced Research and Reviews, 2022, 14(03), 452–459

2. Related work

2.1. Prediction of Land Suitability for Crop Cultivation

Machine learning techniques are used to identify the best crops for a given region, greatly benefiting farmers in crop
prediction. The feature selection (FS) facet of machine learning is a critical component in the selection of relevant
features for a given region, and it keeps the crop prediction process up to date. To choose acceptable features from a
data collection for crop prediction, this paper introduces a novel FS approach called modified recursive feature
elimination (MRFE). Using a ranking algorithm, the suggested MRFE technique chooses and ranks salient features. The
MRFE method selects the most accurate features, while the bagging methodology aids in properly predicting an
appropriate crop, according to the testing results. Various metrics like as accuracy (ACC), precision, recall, specificity,
F1 score, area under the curve, mean absolute error, and log loss are used to assess the performance of the proposed
MRFE approach. The MRFE methodology outperforms other FS methods with 95% accuracy, according to the
performance analysis [2].

2.2. Ascertaining the Fluctuation of Rice Price using Machine Learning

The purpose of this research is to use a machine learning approach to anticipate rice prices in Bangladesh. The price
was predicted using data from Bangladesh's Ministry of Agriculture website. Support Vector Machine (SVM), K-Nearest
Neighbor (KNN), Naive Bayes, Decision Tree, and Random Forest were among the machine learning techniques used to
make this prediction. All of these algorithms are compared to see which one delivers the greatest results. Despite the
fact that all five algorithms performed similarly, the random forest method stood out as the best. As a result, they
employed the Random Forest to forecast rice prices. This projection could lead to a situation in which the
Bangladesh government and people know how much rice they need to plant in order to meet everyone's food needs.
Based on the findings obtained by the mentioned algorithms, they forecasted the price of rice, whether it is reasonable,
low, or exorbitant. The fundamental restriction of their work was a lack of real data, and instead of gathering data from
the entire country, they simply collected data from a small portion of it. This study was hampered by a lack of real data,
and rather than collecting data from across the country, they relied solely on data from Dhaka to create a dataset. The
inclusion of new data in their dataset is a top priority for future work [3].

2.3. Forecasting Crop productivity

This research provides a realistic and user-friendly yield prediction system for Indian farmers, outlining the
shortcomings of current methods and their practical use in yield prediction. A smartphone application is used to link
farmers to the proposed system. Users may utilize a variety of tools in the mobile application to choose a crop. GPS
assists in determining the user’s position. As input, the user enters the area and soil type. Machine learning algorithms
may be used to identify the most lucrative crop list or to forecast crop yields for a user-selected crop. The built-in
predictor technology assists farmers in predicting crop yields. The built-in recommender system helps the user to
explore the various crops and their yields in order to make better informed judgments. Various Machine Learning
techniques, including SVM, Artificial Neural Network (ANN), Random Forest, Multivariate Linear Regression (MLR), and
KNN, are applied and tested on datasets from Maharashtra and Karnataka to forecast crop productivity. With 95%
accuracy, the random forest is the best among the set of typical algorithms used on the given datasets. In addition, the
system recommends the optimal time to apply fertilizers to increase production. Future study will focus on periodically
updating datasets in order to generate reliable predictions, and the processes can be automated. Another feature that
will be implemented is the ability to offer the appropriate fertilizer for each crop [5].

2.4. Crop Yield Estimation

Machine learning (ML) is an important approach for obtaining practical and real-world solutions to crop yield
difficulties. ML can predict a target/outcome from a set of predictors using Supervised Learning. To obtain the required
results, an appropriate function based on a collection of variables must be created, which will map the input variable to
the desired output. Crop yield prediction is predicting a crop's production based on previous data that includes
parameters like temperature, humidity, pH, rainfall, and crop name. The Random Forest method is used to forecast the
best crop yield as an output. In the agriculture field, the crop yield prediction is mostly appropriate. The more increase
in accuracy results in more profit to the crop yield. It will achieve the most accurate crop prediction possible. The
random forest technique is used to find the optimum crop yield model with the fewest number of models. The proposed
method aids farmers in gaining an understanding of the demand for and pricing of various crops. This approach will
cover the widest range of crops, making it extremely valuable for predicting crop yields in the agriculture industry.
Accurate forecasting of different specified crops across different districts will benefit Indian farmers [1].

453
 World Journal of Advanced Research and Reviews, 2022, 14(03), 452–459

3. Proposed system
In our proposed crop recommendation and yield prediction system, we employed various machine learning techniques.
Machine learning a branch that focuses on the use of data and algorithms to imitate the way that humans learn, gradually
improving its accuracy, from that we have used classification and regression algorithms to predict the crop type and
yield produced Quintal/Hectare respectively [1]. The inputs would be state name, district name, season, area, rainfall,
average humidity, and mean temperature. After data preprocessing and data visualization methods, a separate dataset
with past data, other than user input, is given to do the training and testing.

The classification algorithms and regression algorithms, three algorithms each are used to find the best accuracy
calculated and the results are displayed to the user. The classification algorithms, such as logistic regression, decision
tree, and random forest, weigh the input features so that the output separates one class into positive values and the
other into negative values, while the regression algorithms, such as linear regression, decision tree, and random forest,
predict the output values based on input features from the data fed into the system. The best accuracy is ranked based
on the accuracy computed by each of those algorithms, and the results are displayed to the user.

The output is then displayed using flask, a small and lightweight Python web framework that provides essential tools
and capabilities to make web application development simple. We prepared separate model files for the algorithms that
produced the best results for crop type and yield prediction, and used those to display the results in Flask.

Figure 1 System Architecture

4. Module description
Python programming and a variety of libraries is used to build the system. The flask library is used to create the
frontend. The model is trained and tested using a variety of machine learning models. We utilize a supervised machine
learning approach because we have a labeled dataset with a set of crop cultivation features.

The dataset is first gathered. The data is then preprocessed to make it easier to use in training. The data is divided into
two categories: training and testing. Each machine learning model is trained using training data. The accuracy of each
model is then calculated and compared using testing data to discover the best model for the dataset. A user interface is
built which will display the prediction using the stored training data.

4.1. Data pre-processing

The dataset is collected in this initial module with information such as state name, district name, season, crop, area,
rainfall, and so on. Data preparation is carried out in order to convert the raw data into a readable and acceptable format.
This is done to improve the data's quality.

The first step in data preprocessing is data cleaning, which involves locating duplicate values in the dataset, deleting
null values, and removing unwanted values. The data is then categorized before being encoded into numbers so that the

454
 World Journal of Advanced Research and Reviews, 2022, 14(03), 452–459

model can understand and extract valuable information, as machine learning algorithms only take numeric variables. A
categorical data encoding approach is employed for label encoding in this operation.

4.2. Data visualization

Data visualization is used to create a graphical representation of data in order to have a better understanding of the
dataset. The tabular data is converted into a visual representation, such as a scatter plot, bar chart, heatmap plot, and
so on. Hidden data patterns can be discovered by visualizing the data. The data is then examined to see if there are any
clusters, and if so, whether they are linearly separable/overlapped, and so on. We can simply rule out models that aren't
fit for such data based on this preliminary study, and we will implement only such models that are, saving time and
money. This method enables us to quickly comprehend the dataset and construct machine learning models.

A test harness data is created before selecting an appropriate machine learning model. The test harness is the data
against which an algorithm is trained and tested, as well as the performance metric that will be used to evaluate its
performance. The dataset is separated into training and testing datasets for this purpose. 70% of the time is used for
training and 30% for testing. The model will be trained using both input and output data. The input and output of testing
data will be determined by prediction.

4.3. Choosing machine learning models

In this module, Model selection is done, it is the process of selecting one final machine learning model from among a
collection of candidate machine learning models for a training dataset. Before choosing a machine learning model, it is
found which type of the dataset we are having. For our problem statement we are using labeled data so supervised
machine learning algorithms are chosen for training the dataset. Supervised learning uses a training set to teach models
to yield the desired output [1]. This training dataset includes inputs and correct outputs, which allow the model to learn
over time.

For crop recommendation classification supervised machine learning algorithms are used as we can categorize the
output into classes. Classification models have the task of approximating the mapping function from input variables to
discrete output variables. The main goal is to identify which class/category the new data will fall into. For crop yield
prediction regression supervised machine learning algorithms are used as the prediction gives us an output value which
is continuous [4]. Regression is a supervised learning technique which helps in finding the correlation between variables
and enables us to predict the continuous output variable based on the one or more predictor variables.

4.4. Comparing machine learning algorithms

The chosen machine learning algorithms are applied to the prepared data to solve the problem. The training dataset is
applied for each of the algorithms one by one and trained. Using each of these algorithms, crop yield prediction and
prediction of crop name is done. For crop name prediction classification algorithms are used and for crop yield
prediction regression algorithms are used. Then, using testing data evaluation of each model is done. It is important to
compare the performance of multiple different machine learning algorithms consistently. Each model will have different
performance characteristics.

Using resampling methods like cross validation, it is estimated for how accurate each model may be on unseen data.
Then, using various performance metrics the performance of each algorithm is evaluated. A classification report is
generated by finding precision, recall, F1 score [2]. By comparing the results the best algorithm that can be used for this
model is found.

4.5. Storing the trained data

The best machine learning model which gives a higher accuracy for the dataset is selected. Using this model, the dataset
is applied and training is made. The trained data is exported and stored in file for using during the prediction.

4.6. Deployment using flask

A user interface is built which displays crop prediction results. During the prediction the trained data is loaded. Users
can input agricultural parameters. The model which we built will take the input and make predictions. The predicted
result is displayed to the user through the user interface that is built.

455
 World Journal of Advanced Research and Reviews, 2022, 14(03), 452–459

5. Machine learning models

We use supervised machine learning models to build our system. The types of supervised machine learning models used
are:

 Classification algorithms – for Crop recommendation

 Regression algorithms – for Yield Prediction

5.1. Logistic regression

When the goal variable is categorical, a supervised learning approach called logistic regression is applied. The target
variable in Logistic Regression is categorical, therefore we must restrict the range of predicted values. Multiclass logistic
regression is also known as multinomial logistic regression. We used this algorithm to predict more than two classes of
crop types.

5.2. Decision tree

Decision Tree algorithm is a type of machine learning algorithm in which the data is continually split according to a
parameter. The decision tree has an if-else structure, with each node splitting into two or more branches using only one
independent variable. It doesn't matter if the independent variable is categorical or continuous. For categorical
variables, the categories are used to determine the node split and for continuous variables, the algorithm generates
various threshold values that serve as the decision-maker. We applied Decision tree classifier for crop name prediction
and Decision tree regressor yield prediction.

5.3. Random forest

A random forest is a supervised machine learning system that uses decision tree algorithms to construct it.
This algorithm determines the outcome based on decision tree predictions. It predicts by averaging the output of
various trees. The precision of the result improves as the number of trees grows. It reduces dataset overfitting and
improves precision. We applied Random forest classifier for crop name prediction and Random forest regressor yield
prediction.

5.4. Linear regression

Linear regression analysis is a statistical technique for predicting the value of one variable based on the value of another.
The dependent variable is the variable you want to predict. The independent variable is the one we are using to predict
the value of the other variable. We applied linear regression classifier for crop name prediction and Linear regression
yield prediction.

6. Results

Figure 2 Home page

456
World Journal of Advanced Research and Reviews, 2022, 14(03), 452–459

Figure 3 Plantation Crops for specific District

Figure 4 Crop Recommendation Form

Figure 5 Crop Recommendation Result

457
 World Journal of Advanced Research and Reviews, 2022, 14(03), 452–459

Figure 6 Yield Prediction form

Figure 7 Yield Prediction result

7. Conclusion
A crop recommendation and yield prediction system has been developed successfully using Supervised Machine
Learning Approach. The analysis began with data preprocessing and cleaning, followed by exploratory analysis using
an agricultural dataset. After that, we used the dataset to train multiple machine learning models, and made various
evaluation processes to find the best algorithm. As a result of evaluation, we use Decision tree classifier for crop
recommendation and Random Forest regressor for yield prediction as they give best accuracies. Our trained algorithm
can predict crops based on the specified characteristics, as well as crop yield rate. As this system will cover the widest
range of crops, farmers will be able to learn about crops that have never been farmed before and will be able to see a
list of all possible crops, which will aid them in deciding which crop to plant. We intent to help people who are planning
to invest in farming without any prior knowledge about farming and how much they can make a profit out of it. It is also
to help people who are new to farming and try to make their way in learning the practice that has been followed for
generations. Thus, our approach can assist farmers in Tamil Nadu, particularly newcomers, in deciding which crop to
produce by predicting crop and yield based on local climatic circumstances.

The proposed system is developed as a website. In future we may try to develop our system as a mobile application
which makes the user use this application even more user friendly. As a future enhancement we may try to train the
model using neural network algorithms. In our system we are not using neural networks as with the dataset now we
are having good accuracy results with machine learning algorithms. We expect when increasing the features of input

458
 World Journal of Advanced Research and Reviews, 2022, 14(03), 452–459

and increasing the size of the dataset, training the model with neural networks would produce even more efficient
results.

Compliance with ethical standards

Acknowledgments
This research work is done by Anusree M, Divya Lakshmi R and Swetha U under the supervision of Sundari V, under the
Department of Computer Science and Engineering, Meenakshi Sundararajan Engineering College, Chennai, India.

Disclosure of conflict of interest

The authors declare no conflict of interest.

References
[1] Jeevan Nagendra Kumar Y, V Spandana, VS Vaishnavi, K Neha, VGRR Devi. Supervised Machine learning Approach
for Crop Yield Prediction in Agriculture Sector. 5th International Conference on Communication and Electronics
Systems. 2020;736-741
[2] Mariammal G, A Suruliandi, SP Raja, E Poongothai. Prediction of Land Suitability for Crop Cultivation Based on
Soil and Environmental Characteristics Using Modified Recursive Feature Elimination Technique With Various
Classifiers. IEEE Transactions on Computational Social Systems. 2021;8(5):1132-1142
[3] Mehedi Hasan Md, Muslima Tuz Zahara, Mahamudunnobi Sykot, Arafat Ullah Nur, Mohd Saifuzzaman, Rubaiya
Hafiz. Ascertaining the Fluctuation of Rice Price in Bangladesh Using Machine Learning Approach. 11th
International Conference on Computing, Communication and Networking Technologies. 2020;1-5
[4] Rakesh Kumar, MP Singh, Prabhat Kumar, JP Singh. Crop Selection Method to maximize crop yield rate using
machine learning technique. International Conference on Smart Technologies and Management for Computing,
Communication, Controls, Energy and Materials. 2015;138-145.
[5] Shilpa Mangesh Pande, Prem Kumar Ramesh, Anmol Anmol, BR Aishwarya, Karuna Rohilla, Kumar Shaurya. Crop
Recommender System Using Machine Learning Approach. 5th International Conference on Computing
Methodologies and Communication. 2021;1066-1071.

459